IMAGE COMPRESSION METHOD WITH NEGLIGIBLE AND QUANTIFIABLE INFORMATION LOSS AND HIGH COMPRESSION RATIO

Description

TECHNICAL FIELD

[0001] The present disclosure relates to data processing, and more specifically to the encoding and compression of image data.

BACKGROUND ART

[0002] The amount of images produced worldwide is in constant growth. Photographic and video data, by its nature, consumes a large part of available digital resources, such as storage space and network bandwidth. Image compression technology plays a critical role by reducing storage and bandwidth requirements. There is therefore a need for high compression ratios whilst ensuring a quantifiably negligible loss of information.

[0003] Imaging is the technique of measuring and recording the amount of light L (radiance) emitted by each point of an object, and captured by the camera. These points are usually laid-out in a two-dimensional grid and called pixels. The imaging device records a measured digital value d_adc(x, y) that represents L(x, y) , where (x, y) are the coordinates of the pixel. The ensemble of values d_adc(x,y) forms a raw digital image M . Typically M consists of several tens of millions pixels and each d_adc(x, y) is coded over 8 to 16 bits. This results in each raw image requiring hundreds of megabits to be represented, transferred and stored. The large size of these raw images imposes several practical drawbacks: within the imaging device (photo or video camera) the size limits the image transfer speed between sensor and processor, and between processor and memory. This limits the maximum frame rate at which images can be taken, or imposes the use of faster, more expensive and more power consuming communication channels. The same argument applies when images are copied from the camera to external memory, or transmitted over a communication channel. For example:

• transmitting raw "4K" 60 frames per second video in real time requires a bandwidth of 9Gbps; and storing one hour of such video requires 32Tbits.
• transmitting a single 40MPixel photograph from a cube satellite over a 1Mbps link takes 10 minutes.
• backing up a 64GByte memory from a photographic camera over a 100Mbps Internet connection takes more than one hour.

[0004] To alleviate requirements on communication time and storage space, compression techniques are required. Two main classes of image compression exist:

A. Lossless image compression, where digital image data can be restored exactly, and there is zero loss in quality. This comes at the expense of very low compression ratios of typically 1:2 or less.

B. Lossy image compression, where the compressed data cannot be restored to the original values because some information is lost/excised during the compression, but which can have large compression ratios of 1:10 and higher. These compression ratios however come at the expense of loss of image quality, i.e. distortion, and the creation of image compression artifacts, i.e. image features that do not represent the original scene, but that are due to the compression algorithm itself.

[0005] Lossy image compression methods, such as those in the JPEG standard or used in wavelet compression, create distortion and artifacts. Distortion is the degree by which the decoded image Q , i.e. the result of compressing and decompressing the original image, differs from the original image M , and is usually measured as the root-mean-square of the differences in the values of each corresponding pixel between Q and M . Lossy compression typically also introduces artefacts, which are a particularly bad type of distortion that introduces image features in Q , i.e. correlations between different pixels, that are not present in M. Examples of artefacts are the "block" artifacts generated by block image compression algorithms like JPEG, but also ringing, contouring, posterizing. These artifacts are particularly nefarious, as they can be mistaken for image features. Lossy image compression algorithms, as described for example in prior art documents [1] and [2] cited below, typically consist of several steps. First, the image is transformed into a representation where the correlation between adjacent data point values is reduced. This transformation is reversible, so that no information is lost. For example, this step would be similar to calculating the Fourier coefficients, i.e. changing an image that is naturally encoded into a position space, into a special frequency space. The second step is referred to as quantization. This step truncates the value of the calculated coefficients to reduce the amount of data required to encode the image (or block, consisting of e.g. 16x16 pixels). This second step is irreversible, and will introduce quantization errors when reconstructing the image from the aforementioned coefficients. The quantization error will cause errors in the reconstructed image or block. Predicting or simulating what the error would be for any image or block is technically impossible, due to the extremely large number of values that such image or block may take. For example, a 16x16 block, with 16-bits per pixel, may take 2^16×16×16≅10⁴⁹³² different values, impossible to test in practice. If the transformation is not restricted to blocks, but rather is a function of the entire image, the number of possible values becomes considerably larger.

[0006] More particularly, in standard lossy image compression techniques (e.g. document [1]), a lossless "image transformer" is first applied to the image, resulting in transformed image, where each data-point is a function of many input pixels and represents the value of a point in a space that is not the natural (x,y) position space of the image, but rather is a transformed space, for example a data-point might represent a Fourier component. A lossy "image quantizer" is applied on the transformed image as a second step. The amount of information loss cannot be accurately quantified in standard lossy image compression techniques for the above mentioned reason of the extremely large number of values that such image or block may take and is additionally complicated by the fact that the lossy operation is applied on a space that is not the natural image space.

[0007] Attempts have been made at characterizing the quality of the output of image compression algorithms. These efforts have typically focused on characterizing the quality of the reconstructed compressed image with respect to the human visual system, as reviewed in documents [2, 3, 4, 5] described below. Several metrics have been devised to characterize the performance of the compression algorithm, however, these metrics have significant drawbacks. One such drawback is that as they relate quality to the human visual system and human perception, which are highly dependent on the subject viewing the image, on the viewing conditions (e.g. eye-to-image distance, lighting of the image, environment lighting conditions, attention, angle, image size) as well as on the image rendering algorithms (e.g. debayering, gamma correction) and characteristics of the output device, such as display, projector screen, printer, paper etc. A second drawback of characterizing the quality of image compression algorithms with respect to the human visual systems or models thereof is that for such methods to be relevant, no further image processing must take place after image compression, as such processing would make some unwanted compression artifacts visible. For example, a compression algorithm might simplify dark areas of an image, removing detail, judging that such detail would not be visible by the human eye. If it is later decided to lighten the image, this detail will have been lost, resulting in visible artifacts. Yet another problem of the above-mentioned methods is that they are unsuitable for applications not aimed solely at image reproduction for a human observer, as for example are images for scientific, engineering, industrial, astronomical, medical, geographical, satellite, legal and computer vision applications, amongst others. In these applications data is processed in a different way as by the human visual system, and a feature invisible to the untrained human eye and therefore removed by the above-mentioned image compression methods, could be of high importance to the specific image processing system. For example, inappropriate image compression of satellite imagery could result in the addition or removal of geographical features, such as roads or buildings.

[0008] In this context, one may also mention document EP 1 215 909 (K. Kagechi et al.) which is directed to providing an image encoding - and decoding device allowing the maximum compression rate whilst guaranteeing a visually uniform level of picture quality. Therefore, this device aims to realize a method of automatic setting of image compression parameters reflecting the characteristics of human vision by extracting the pixel regions of the image which are felt having deteriorated visually and either to utilise only the degree of distortion thereof for the evaluation of picture quality or by classifying the image blocks according to their properties to set up a separate criterion for evaluation. Therefore, compression parameters are varied in this device for each block of an image by calculating its characteristic distortion using the original - and the encoded-decoded image, such that, if the corresponding device allows to improve visual image quality whilst increasing the compression rate, it doesn't provide for a well quantifiable information loss, given that the latter varies in each block of an image.

[0009] More recently, attempts at a more quantitative approach on the information loss in digital image compression and reconstruction have been made [6], in particular trying to examine information loss given by the most common image compression algorithm: Discrete Wavelet Transform (DWT), Discrete Fourier Transform (DFT) and 2D Principal Component Analysis (PCA). Several "quantitative" metrics are measured on several images compressed with several algorithms, for example quantitative measures of "contrast", "correlation", "dissimilarity", "homogeneity", discrete entropy, mutual information or peak signal-to-noise ratio (PSNR). However, the effect of different compression algorithms and parameters on these metrics highly depends on the input image, so that it is not possible to specify the performance of an algorithm, or even to choose the appropriate algorithm or parameters, as applying them on different images gives conflicting results. A further drawback of the methods described in [6], is that although these methods are quantitative in the sense that they output a value that might be correlated with image quality, it is unclear how this number can be used: these numbers cannot be used as uncertainties on the value, the compression methods do not guarantee the absence of artifacts, and it is even unclear how these numbers can be compared across different compression methods and input images. Also, these quantitative methods do not distinguish signal from noise, so that for example, a high value of entropy could be given by a large amount of retained signal (useful) or retained noise (useless).

[0010] The impossibility for the above-mentioned methods to achieve compression with quantifiable information loss arises from the fact that the space in which the algorithm looses data is very large (e.g. number of pixels in the image or block times the number of bits per pixel), and cannot therefore be characterized completely, as we have mentioned earlier.

[0011] Image-processing systems that act independently on individual (or small number of) pixels have been known and used for a long time. These systems typically provide a "look-up table" of values so that each possible input value is associated with an "encoded" output value. The main purpose of such processing has been to adapt the response curves of input and output devices, such as cameras and displays respectively. The functions used to determine the look-up tables have typically been logarithmic or gamma law, to better reflect the response curve of the human visual system as well as the large number of devices designed to satisfy the human eye, starting from legacy systems like television phosphors and photographic film, all the way to modern cameras, displays and printers that make use of such encodings. Although not their primary purpose, these encodings do provide an insignificant level of data compression, such as for example encoding a 12-bit raw sensor value over 10-bits, thus providing a compression ratio of 1.2 : 1, as described, for example in [7].

[0012] Finally, one may also mention in this context another category of disclosures, like document US 2010/189180 (M. Narroschke et al.) which discloses a method for coding a video signal using hybrid coding in order to realize enhanced quantization as compared to prior art hybrid video coding. To achieve the latter, the method provides an adaptive control allowing to switch between the frequency - and the spatial domain for transforming the prediction error signal which is generated in a previous step by subtracting a motion compensation prediction signal of the video signal from the input video signal in order to reduce temporal redundancy in the video signal. The output signal of the quantization step occurring either in the frequency - or in the spatial domain is then passed to an entropy encoder which provides the final output signal to be transmitted or stored. Said adaptive control allows for deciding for each block of the video image whether it is to be coded in the frequency - or in the spatial domain and generates a side information on the domain effectively used. This amounts to enhancing quantization as compared to prior art hybrid video coding by switching, for each image block to be quantized, between the frequency - and the spatial domain, depending on where the rate distortion costs are lowest and thus efficiency is highest, this being decided via said adaptive control. In turn, the rate distortion costs, i.e. the information loss, is varying at the level of quantization of each image block, such that quantifying the information loss generated by this method is difficult.

REFERENCES

[0013] 

[1] W.M. Lawton, J.C. Huffman, and W.R. Zettler. Image compression method and apparatus, 1991. US Patent 5,014,134.

[2] V.Ralph Algazi, Niranjan Avadhanam, Robert R. Estes, and Jr. Quality measurement and use of pre-processing in image compression. Signal Processing, 70(3):215 - 229, 1998.

[3] Ahmet M Eskicioglu and Paul S Fisher. Image quality measures and their performance. Communications, IEEE Transactions on, 43(12):2959-2965, 1995.

[4] Philippe Hanhart, Martin Rerabek, Francesca De Simone, and Touradj Ebrahimi. Subjective quality evaluation of the upcoming hevc video compression standard. 2010.

[5] Weisi Lin and C.-C. Jay Kuo. Perceptual visual quality metrics: A survey. Journal of Visual Communication and Image Representation, 22(4):297 - 312, 2011.

[6] Zhengmao Ye, H. Mohamadian, and Yongmao Ye. Quantitative analysis of information loss in digital image compression and reconstruction. In Communication Software and Networks (ICCSN), 2011 IEEE 3rd International Conference on, pages 398-402, May 2011.

[7] Lars U. Borg. Optimized log encoding of image data, 2012. US Patent 8,279,308.

[8] T.A. Welch. A technique for high-performance data compression. Computer, 17(6):8-19, June 1984.

[9] I. Matsuda, H. Mori, and S. Itoh. Lossless coding of still images using minimum-rate predictors. In Image Processing, 2000. Proceedings. 2000 International Conference on, volume 1, pages 132-135 vol.1, 2000.

SUMMARY

[0014] In view of the above, it is therefore an object of the invention to provide a data compression method that permits to compress data with a high compression ratio with minimal information loss and to provide a technique for determining the information loss or the uncertainty with respect to the compression ratio.

[0015] A further object of the invention is to provide a decompression method adapted to decompress the data compressed by the compression method of the invention.

[0016] A further object of the invention is a data processing method comprising the compression method and the decompression method.

[0017] Here are described the embodiments of an image compression system and method that have a number of advantages:

• Precise estimation of information loss at each chosen compression ratio.
• A specific, quantifiable, bound for the compression ratio at which information loss remains negligible.
• High compression ratio (e.g. 8:1) with negligible information loss.
• No compression artifacts.
• Both the pixel value and uncertainty ("error bar") on the pixel value can be recovered from the compressed image data and used in subsequent processing steps.
• Allows the user to choose the desired compromise between information loss and level of compression.
• The system can be characterized for all possible photographs, so that the amount of information lost can be predicted for any possible photograph.

[0018] To achieve the above advantages, the embodiments presented here distinguish the amount of image information from the amount of image data. Information is the useful knowledge acquired about the photographed object by means of the camera. Data is the digital representation of the measurement result, and contains information as well as noise and redundancy. In general, the amount of data captured will be much larger than the amount of information, so that it is possible to eliminate a large amount of data whilst eliminating only very little information.

[0019] We give an overview of an embodiment, before going into its details in the next section. Referring to figure 4, first, the imaging system and compression method (e.g. sensor and processing) are characterized, for example by calibration, by simulation or theoretically, to determine the optimal pre-compression parameters. In contrast to the standard lossy image compression techniques, we apply an appropriately-tuned lossy pre-compression, or quantizer directly on the acquired raw image data, as a pre-compression step. The amount of information lost by this quantizer can be calculated or measured, and appropriate parameters chosen to guarantee that such information loss remains negligible. Pre-compression removes most of the noise, which cannot, by its nature, be losslessly compressed, however, thanks to the properly chosen parameters, this step only removes a negligible amount of information. The appropriately-quantized image data, that now contains mostly information and little noise, is fed to a lossless compression algorithm as a second step. This algorithm will work very effectively, as more redundancy can be found in this clean data than could be found in the raw image sensor data that contained a large amount of noise. As now the second step is lossless, it will not reduce the amount of information present in the image and therefore does not need to be characterized for information loss.

[0020] We characterize the amount of information lost as the increase in the uncertainty (i.e. the standard deviation, or "error bar") in the measured value of the radiance after the application of the compression and decompression algorithm, and define that this loss is negligible when the uncertainty on pixel values after decompression remains smaller than the standard deviation between the pixel values of consecutive photographs of the same subject.

[0021] The invention relates to a data compression method wherein the data comprises noise and information, comprising a data acquisition step, a pre-compression parameter selection step wherein said pre-compression parameters are linked to an information loss, a compression step, and a storage step, said compression step comprises a lossy pre-compression for removing some noise of the data, carried out using the pre-compression parameters determined in the calibration step followed by lossless compression for compressing the remaining data.

[0022] Preferably, the compression method comprises a parameter determination step for determining pre-compression parameters by establishing a plot of the information loss vs. said pre-compression parameters.

[0023] The parameter determination step comprises a system calibration step outputting the amount of image information that is lost by the compression algorithm as a function of pre-compression parameters.

[0024] Preferably, the system calibration step comprises

• setting the amount of light L to be measured, starting from L=0 (800),
• acquiring a number of calibration images (801) through said camera comprising an image sensor (306), an amplifier (310) and a converter (312),
• calculating an input uncertainty σ_i(L)/L on the calibration data comprising said number of calibration images (801),
• setting a set of pre-compression parameters g,
• applying a lossy pre-compression followed by a lossless compression to the calibration data;
• decompress the calibration data;
• calculate output uncertainty on the calibration data;
• repeat the above steps by increasing L and g, until L_max and g_max have been reached,
• report information loss as a function of the set of pre-compression parameters g.

[0025] The invention also relates to a data decompression method according to independent claim 6.

[0026] The invention also relates to a data processing method comprising the compression method followed by the decompression method described above.

[0027] The invention also relates to a computer program run on a computer or a dedicated hardware device, like a FPGA, adapted to carry out the data processing method described above.

DETAILED DESCRIPTION

Imaging system

[0028] We start by describing how an image is typically acquired more in detail. This is illustrated in figure 2. An imaging device aims at measuring (estimating) the amount of light emitted by a specific area (302) of an object. We call this amount of light the radiance L(x, y) illuminating a pixel (308) with coordinates (x, y) on the image sensor (306). Light from this area is collected by an imaging system (304), such as a camera objective, and focused on a specific pixel (308) of the image sensor (306). During the camera's exposure time T , a number of photons γ(x, y) will impinge on the pixel (308), and generate a number of photoelectrons n(x,y). These electrons (a charge) will typically be translated to a voltage v(x, y) by an amplifier (310). This voltage is digitized by an analogue-to-digital converter (ADC) (312) that outputs a digital value d_ADC(x, y). The raw image data (314, 212) is the ensemble of values d_ADC(x, y) for all (x, y) comprising the image. From d_ADC(x, y) it is possible to estimate L_σi(x, y), i.e. the value of the radiance L(x, y) with an uncertainty σ_i(x, y) . The uncertainty is given by noise, such as shot noise and read noise. If the sensor, amplifier and ADC are linear, the mean value 〈d_ADC(x, y)〉 can be related to the mean value of the number of photoelectrons 〈n〉 by proportionality constant ζ and an offset constant c :

[0029] Similarly, the mean number of photoelectrons 〈n〉 is proportional to the number of photons γ(x, y) that impinge on the pixel during the integration time T with a proportionality constant called the quantum efficiency QE :

[0030] The number of photons γ(x, y) is proportional to the radiance L(x, y) , i.e. the amount of luminous power emitted by the object by unit area and solid angle. γ(x, y) is also proportional to other constants, such as the observed area A (function of the imaging system (304) and pixel size), observed solid angle Q (function of the imaging system (304)), exposure time T , and inversely proportional to photon energy E=hc/λ where h is Planck's constant and λ is the wavelength of the incident light. This can be summarized as:

[0031] Substituting Equation (EQ3) into Equation (EQ2) and the resulting equation into Equation (EQ1), one obtains:

[0032] Setting Z=ζQEAΩT/E ,

[0033] So that L(x, y) can be directly evaluated from 〈d_ADC(x, y)〉 as:

[0034] And on for a single shot (i.e. without averaging) one may evaluate L_σi(x, y) as:

[0035] Equation (EQ7) serves as an example for a system that is very linear. If this is the case, and the sensor, amplifier and converter result in a shot-noise limited measurement, the number of photoelectrons will be Poisson-distributed, so that the uncertainty in photoelectron number will be

, and the relative uncertainty

.

[0036] If the sensor, amplifier or ADC presents non-linearity, it is necessary to calibrate them, for example by creating a look-up-table LUT_d→L and LUT_d→σi that associates a value of L and σ_i, to each possible value of d_ADC . Below we will detail how this is done in practice.

[0037] In conclusion, the raw data will be the recorded digital value d_adc(x, y) , the information acquired about the object radiance L(x, y) will have value L_σi(x,y) with uncertainty σ_i(x, y) . When image data undergoes compression and decompression, resulting in the possibility to infer an output L_σo(x, y) with an uncertainty σ_o(x, y) , it is then possible to quantify the amount of information lost by looking at the relative increase in uncertainty: σ_o(x,y)/σ_i(x, y) , i.e. the relative increase in error bars.

[0038] An overview of a preferred embodiment of the method of the present invention will now be described with reference to figure 1.

[0039] This preferred embodiment is a data processing method which is preferably split into 3 phases as depicted in figures 4A, 4B and 4C.

1) The first phase also called "parameter determination" (250) and represented in figure 4A serves to determine pre-compression parameters (210) that achieve a high compression ratio while guaranteeing that the information lost is below a specific bound (208).

2) The second phase is the compression method of the present invention represented in figure 4B which compresses image data using the parameters (210) determined in the parameter determination phase (250). The compressed data are transmitted and/or stored (220).

3) The third phase is the decompression method of the present invention represented in figure 4C which recovers compressed data (222) from the transmission channel or storage (220), as well as the bound on information loss (224). This information is used to obtain decompressed data (228), as well as the uncertainty on this data (230). Both of these can be used for processing (232) and display or other usage (234).

[0040] It is important to note that even if the data processing method of the invention comprises these three phases, each single phase can be carried out independently from each other. Such that the present invention comprises a parameter determination method, a compression method, a decompression method as well as a data processing method comprising one or more of these three methods.

[0041] We will now describe the first phase of parameter determination.

Parameter determination phase for system calibration

[0042] As shown in figure 4A, in order to carry out a system calibration allowing to determine the pre-compression parameters and the bound on the information loss, first, the camera system is characterized (200) to determine the amount of image information that is lost (202) by the compression algorithm as a function of pre-compression parameters. This characterization is described in detail here below with reference to figure 3. In a simple exemplary embodiment, that works well for shot-noise limited, linear systems, the pre-compression parameter will be a single variable g taking values between 1 (no compression and data loss) and infinity (high compression and all data lost). This parameter will be used in the pre-compression algorithm described in the compression phase detailed below.

[0043] A sample output of this step is shown in figure 6 representing a plot of the uncertainty/information loss with respect to parameter g. With this data (202) in hand, and according to the preferences of the user (204), appropriate compression parameters are chosen. In the instance of figure 6, for example, the pre-compression parameter g can be increased up to a value of 2, with the amount of information lost remaining below the negligible loss bound. Figure 5 shows the quantified uncertainty increase (information loss) bound (208) for values of L between 0 and 500 (here the units are normalized to the mean number of photoelectrons 〈n〉 ).

[0044] In the description below, the pre-compression parameters is a single parameter relating to a pre-defined pre-compression function; however, in general, it may be represented by multiple parameters, or the look-up-tables LUT_d→L and LUT_d→σi . The selected parameters and associated uncertainty increase bounds are saved to be reused in the next phase: Compression.

System characterization

[0045] An embodiment of this procedure is shown in Figure 3. System characterization evaluates the amount of information lost by the compression system as a function of compression parameter and input radiance L .

[0046] For clarity reason, a list of the symbols used here is available at the end of the present specification.

[0047] First, the amount of light to be measured is set, starting from L=0 (800), then, a number of images is acquired (801). This can be done using the real image acquisition system, by simulating such a system, or by theoretical modeling. In this example, we simulate the data arising from a shot-noise limited linear sensor. Assuming linearity means that the number of photoelectrons n will be proportional to the radiance L to be measured (at constant T, Ω, QE, A, E ): 〈n〉=LZ/ζ . For simplicity of the simulation, we set ζ=1 and c=0 so that d_ADC=n . For a shot-noise limited system, n will be distributed according to a Poisson distribution N(LZ/ζ)=N(〈n〉) that has mean 〈n〉 , variance 〈n〉 and standard deviation

. With the above simplifications, the relative error σ_i/〈L_i〉 will be equivalent to the relative uncertainty in the measurement of the number of photoelectrons

. To simulate sensor data in (801) for a specific value of 〈n〉 , we generate a number of samples n_j issued from N((n)) . Instead of simulating data, (801) may acquire data from the imaging system.

[0048] The relative input uncertainty σ_i(L)/L is calculated or numerically computed as root-mean-square (RMS) difference between each L_i , the measured value of L , and the actual (setpoint) value of L , divided by L itself (802).

[0049] A pre-compression parameter, or set of pre-compression parameters is chosen (804). In our simple example, we use a single parameter g . However, more in general, this could be a look-up-table (LUT) associating an output value of L_o to each possible value of L_i .

[0050] Lossy pre-compression is then applied to each input measured value of L_i (808). In this embodiment, we obtain the pre-compressed digital value d_c from the function

where n is the number of photoelectrons on that sample, i.e.

or, with respect to the ADC output, n= (d_ADC- c)/ζ . The double-line "brackets" symbolize taking the nearest integer. In this case, g represents the reduction factor in the number of bits required to encode each possible number of photoelectrons. For example, encoding a number of photoelectrons between 0 and 2¹⁶-1=65535 requires 16 bits. For g=1 , no pre-compression would arise, and 16 bits would be required. For g=2 , the number of encoding bits is reduced by a factor of 2, so that the result of the pre-compression can be encoded over 8 bits. More in general, any LUT can be applied.

[0051] A lossless compression algorithm may then be applied (810), either on a pixel-by-pixel basis, or on larger blocks, or on the entire pre-compressed image. We call the data to which this lossless compression is applied F . As after the lossy pre-compression is applied, the amount of data is reduced, but also, the entropy of such data is reduced (noise is removed) so that lossless compression will be very effective. In our example, we have used several algorithms, ranging from the universal and simple Lempel-Ziv-Welch (LZW) [8] to image-content-based algorithms such as minimum-rate predictors (MRP) [9]. The algorithm here can be chosen at will, as being lossless, it will not lose any information, and will not affect the uncertainty of the image. In some instances, if compression speed is of the essence, this step may be a simple "pass-through" not affecting the pre-compressed data at all.

[0052] In the next step (812), the output f of this lossless compression is decompressed to obtain F , then, the inverse function of the lossy compression is applied (or the inverse LUT). For example, the inverse of

is d_o → d_c^g .

[0053] From the de-compressed output d_o it is possible to calculate a value L_o for the radiance, e.g. L_o=d_oζ/Z . In (814) we calculate the error that was introduced as the difference L_o - L_i . This error can be estimated for several samples of the acquired image data, and the root-mean-square (RMS) of this error calculated as σ_o(L, g) , and compared to the input error (uncertainty) σ_i(L) to obtain a well-defined, properly quantified, relative increase in uncertainty Δ(L, g)=σ_o(L, g)/σ_i(L) . The compression ratio r can also be calculated as the size of the original image data divided by the size of output f of the lossless compression.

[0054] By repeating the above process for a number of image acquisition samples, for a number of compression factors g (816, 824) and for a number of Radiances, integration times or generated photoelectrons (818, 822), one obtains a characterization of the compression system (820), i.e. the amount of information lost Δ(L,g) as a function of pre-compression parameters g and pixel radiance L (or equivalently, photoelectron number n or pixel value d_c or d_o , as all these are relate by known equations).

[0055] Sample results of this characterization are shown in figures 6, where g is varied at fixed L , and 5, where L is varied such as to produce between zero and 500 photoelectrons 〈n〉 , and the pre-compression is held fixed at g =2. The results show that Δ(L,g)=σ_o(L,g)/σ_i(L)=1.15 across the range for all these settings. This is smaller than what we define as the bound for negligible information loss, i.e. the RMS of the difference between consecutive measurements ("photographs") of the same pixel, of exactly the same subject taken with an ideal shot-noise-limited sensor, that is

.

[0056] Testing this pre-compression factor g=2 , and therefore Δ=1.15 on an image of a 10 EUR note, acquired with an ATIK383L scientific camera, the compression factors were as follows:

Original image file: 5,164,276 bytes

Lossless compressed original file (ZIP): 4,242,951 bytes.

Lossy Pre-compressed image: 2,580,219 bytes

Lossless compressed pre-compressed image (ZIP): 1,015,808 bytes

Lossless compressed pre-compressed image (MRP [9]): 544,336 bytes

[0057] To give a compression ratio of 9.5:1 and increasing the uncertainty by a factor of only 1.15. Figure 5 plots both the error-bars with the original uncertainty σ_i and with the, barely distinguishably larger, output uncertainty σ_o , from which it is evident that this uncertainty increase is negligible.

[0058] It has to be noted that, when keeping parameters A,Ω,T,E constant, shot noise will manifest itself with the number of photoelectrons n following a Poisson distribution with mean and variance 〈n〉 and with standard deviation

.

[0059] Once, the system characterization is achieved and the pre-compression parameters are available, the user can select the required parameters and continue with the compression phase.

Compression method

[0060] An embodiment of this compression method is illustrated in figure 4B (252). Pre-compression parameters (210) g and associated uncertainty increase bounds (208) Δ are taken from previous phase (250) "parameter determination". In this example, g=2, Δ=1.15 , however it may be different depending on the parameter estimation results. Raw image data is first acquired (212, 314), either from the image processing pipeline or from disk/memory. Then, a lossy pre-compression is applied (214) in the same way as described in step 808 of the section "System characterization", using pre-compression parameter g (210). Pre-compressed data is then losslessly compressed in step (216) such as to yield compressed image data (218) as described in step 810 of the section "System characterization". Compressed data is then stored or transmitted (220). The bound on information loss (208) may also be included into the stored/transmitted data for future use, or may be a pre-defined parameter for the system.

[0061] Once, the compression is achieved and the compressed data are stored or transmitted (that actually can be considered as a mobile storage in a channel), the user can decide to decompress the data to recover the data and therefore pass to the last phase that is the decompression method.

De-compression method and usage

[0062] An embodiment of this method is illustrated in figure 4C (254) "Decompression and usage". First, the compressed data (222) are obtained from the transmission link or storage (220). Ideally, these data were compressed by the above compression method. The bound on information loss (224) is also obtained from the transmission link or storage, but could also be a pre-defined part of the system. This information is fed into a de-compression routine (226) that has already been described in step 812 of the section "System characterization". This step will output decompressed data (228), and the uncertainty on such de-compressed data (230).

[0063] As a matter of fact, using the example of the compression and decompression functions described in section "System characterization", the decompressed image data uncertainty (230) is calculated as

where d_o is the decompressed data output for that pixel and Δ is preferably 1.15.

[0064] This data (228) and, optionally, the associated uncertainty (230), can then be used to effectively process the image (232), resulting in data to be output to display, printing, or further analysis or processing (234).

Table of symbols

Symbol	Description
L	Radiance of each point of an object
L(x,y)	Radiance of a point of the object corresponding to pixel coordinates (x,y)
Li(x, y)	Estimate of radiance L (x,y) as derived from measurement, before compression.
σi(x,y)	Uncertainty (standard deviation) associated with the measurement estimate of the radiance L(x, y) at point (x, y), before compression.
Lo(x, y)	Estimate of radiance L (x,y) as derived from measurement, after compression and decompression.
σo(x, y)	Uncertainty (standard deviation) associated with the measurement estimate of the radiance L(x, y) at point (x, y), after compression and decompression.
γ(x, y)	Number of photons impinging on pixel (image sensor element) at coordinates (x,y)
QE	Quantum efficiency, i.e. the probability that a photon impinging on a pixel is converted into a photoelectron.
n(x, y)	Number of photoelectrons generated at pixel with coordinates (x, y)
v(x, y)	Voltage
dadc(x, y)	Digital number resulting from the digitization of the voltage generated by the photoelectrons from the pixel at coordinates (x,y)
M	Ensemble of captured digital values dadc(x, y), forming a raw digital image, before compression
Q	Ensemble of pixel values after they have undergone compression and decompression.
R , r	Compression ratio, i.e. size of the storage space required to store the input image divided by the size of the storage space required to store the compressed image.
T	Exposure time during which the image sensor absorbs photons and converts them to photoelectrons.
ζ	Factor relating the output
	dadc(x, y) of the analog-to-digital converter to the number of photoelectrons .
c	Offset constant relating the output
	dadc(x, y) of the analog-to-digital converter to the number of photoelectrons n(x,y).
A	Area observed by the pixel, this is a function of optics of the camera, and remains constant for our purposes.
Ω	Observed solid angle, also a function of the optics of the camera and just a constant in our treatment.
E	Photon energy
h	Plack's constant
c	Speed of light
Z	A constant Z=ζQEAΩT/E
Δn	Uncertainty (standard deviation) in the number n of photoelectrons.
g	Set of pre-compression parameters or single pre-compression parameter
N(µ)	Poisson distribution of mean µ
nj	Number of samples
dc	Pre-compressed digital value
do	De-compressed digital value
f	Compressed file, after lossless compression
F	De-compressed file
Δ(L,g), Δ	Relative increase in uncertainty, here as a function of light hitting the pixel and pre-compression parameter, and on it's own.
〈x〉	Represents the mean of x
Lmax	Maximum value of L to be tested when iterating
gmax	Maximum value of g to be tested when iterating

Claims

1. Data compression method wherein said data comprises noise and information, comprising

- an image data acquisition step performed by a camera in order to obtain image data to be compressed,

- a parameter determination step allowing to determine pre-compression parameters which are linked to an information loss,

- a compression step, and

- a storage step,
said compression step comprising

- a lossy pre-compression for removing some noise of the image data followed by

- a lossless compression for compressing the remaining image data,
characterized in that

- said parameter determination step comprises

∘ a system calibration step using calibration data comprising a number of calibration images each acquired by or simulated for said camera at known radiance of light and outputting the amount of image information that is lost by the compression step as a function of pre-compression parameters, said information loss being represented by the relative increase in the uncertainty associated with the measurement estimate of the radiance of light of said image data after said compression step and a decompression step as compared to the uncertainty of said image data before said compression step, and

° a pre-compression parameter selection step allowing to choose said pre-compression parameters,

- said lossy pre-compression being carried out using the pre-compression parameters selected in the pre-compression parameter selection step.

2. Data compression method of claim 1, characterized in that said parameter determination step determines said pre-compression parameters by establishing a plot of the information loss vs. said pre-compression parameters.

3. Data compression method of one of claims 1 to 2, characterized in that the parameter determination step determines a bound on information loss and pre-compression parameters guaranteeing that the information loss is below said bound on information loss.

4. Data compression method of one of claims 1 to 3, characterized in that the system calibration step comprises

- setting the radiance of light L(x, y) to be measured, starting from L=0 (800),

- acquiring a number of calibration images (801) through said camera comprising an image sensor (306), an amplifier (310) and a converter (312),

- calculating an input uncertainty σ_i(x, y) on the calibration data comprising said number of calibration images (801),

- setting a set of pre-compression parameters g,

- applying a lossy pre-compression followed by a lossless compression to the calibration data,

- decompressing the calibration data,

- calculating an output uncertainty σ_o(x, y) on the calibration data,

- repeating the above steps by increasing L and g until L_max and g_max have been reached,

- report information loss as a function of the set of pre-compression parameters g,

where

L(x, y) is the radiance of light at point (x, y),

g is the set of pre-compression parameters,

σ_i(x, y) is the uncertainty associated with the measurement estimate of the radiance L(x, y) at point (x, y) , before compression,

σ_o(x, y) is the uncertainty associated with the measurement estimate of the radiance L(x, y) at point (x, y) , after compression and decompression.

5. Data compression method of one of claims 1 to 4, characterized in that the lossy pre-compression of the image data is applied in the natural, non-transformed space of the image data.

6. Data decompression method comprising

- obtaining compressed image data that have been compressed by the compression method of one of claims 1 to 5,

- obtaining a bound on information loss associated with said compressed image data and stored or transmitted by the compression method of one of claims 1 to 5,

- decompressing the compressed image data,

- determining an uncertainty on the decompressed image data according to the following relation:

where:

Δ(g, L) is the relative increase in uncertainty as a function of the radiance of light L hitting the pixel and the set of pre-compression parameters g, obtained during the system calibration step of the compression method of one of claims 1 to 5,

σ_o(x, y) is the uncertainty associated with the measurement estimate of the radiance L(x, y) at point (x, y) , after compression and decompression,

σ_i(x, y) is the uncertainty associated with the measurement estimate of the radiance L(x, y) at point (x,y) , before compression, obtained during the system calibration step of the compression method of one of claims 1 to 5,

L_o(x,y) is the estimate of radiance L (x, y) as derived from measurement, after compression and decompression,

L_i(x,y) is the estimate of radiance L (x, y) as derived from measurement, before compression, obtained during the system calibration step of the compression method of one of claims 1 to 5,

g is the set of pre-compression parameters obtained during the system calibration step of the compression method of one of claims 1 to 5.

7. Data decompression method of claim 6, characterized in that g is 2 and Δ(g,L_o) is 1.15.

8. Data processing method comprising the compression method of one of claims 1 to 5 followed by the decompression method of one of claims 6 to 7.

9. Computer program run on a computer or a dedicated hardware device, for example a FPGA, adapted to carry out the data processing method of claim 8.

Ansprüche

1. Datenkompressionsverfahren, wobei die besagten Daten Rauschen und Information umfassen, umfassend

- einen Bilddatenerfassungsschritt, der von einer Kamera durchgeführt wird, um zu komprimierende Bilddaten zu erhalten,

- einen Parameterbestimmungsschritt, der es erlaubt, Vorkompressionsparameter zu bestimmen, die mit einem Informationsverlust verbunden sind,

- einen Kompressionsschritt und

- einen Speicherschritt,
wobei der besagte Kompressionsschritt

- eine verlustbehaftete Vorkompression zum Entfernen von etwas Rauschen aus den Bilddaten gefolgt von

- einer verlustfreien Kompression zum Komprimieren der verbleibenden Bilddaten aufweist,
dadurch gekennzeichnet, daß

- der besagte Parameterbestimmungsschritt

∘ einen Systemkalibrierungsschritt, welcher Kalibrierungsdaten, die eine Anzahl von Kalibrierungsbildern umfassen, die jeweils bei bekannter Lichtstrahlung von der besagten Kamera erfaßt oder für diese simuliert wurden, verwendet und welcher die Menge an Bildinformation ausgibt, die durch den Kompressionsschritt als Funktion der Vorkompressionsparameter verloren geht, wobei der besagte Informationsverlust durch die relative Zunahme der Unsicherheit, die mit der Meßschätzung der Lichtstrahlung der besagten Bilddaten nach dem besagten Kompressionsschritt und einem Dekompressionsschritt verbunden ist, im Vergleich zu der Unsicherheit der besagten Bilddaten vor dem besagten Kompressionsschritt dargestellt wird, und

∘ einen Vorkompressionsparameterauswahlschritt, welcher es erlaubt, die besagten Vorkompressionsparameter auszuwählen, aufweist,

- wobei die besagte verlustbehaftete Vorkompression unter Verwendung der im Vorkompressionsparameterauswahlschritt ausgewählten Vorkompressionsparameter ausgeführt wird.

2. Datenkompressionsverfahren nach Anspruch 1, dadurch gekennzeichnet, daß der Parameterbestimmungsschritt die besagten Vorkompressionsparameter bestimmt, indem ein Diagramm des Informationsverlusts in Abhängigkeit von besagten Vorkompressionsparametern erstellt wird.

3. Datenkompressionsverfahren nach einem der Ansprüche 1 bis 2, dadurch gekennzeichnet, daß der Parameterbestimmungsschritt eine Informationsverlustgrenze und Vorkompressionsparameter, die garantieren, daß der Informationsverlust unter der besagten Informationsverlustgrenze liegt, bestimmt.

4. Datenkompressionsverfahren nach einem der Ansprüche 1 bis 3, dadurch gekennzeichnet, daß der Systemkalibrierungsschritt umfaßt

- Einstellen der zu messenden Lichtstrahlung L(x, y) ausgehend von L=0 (800),

- Erfassen einer Anzahl von Kalibrierungsbildern (801) mittels der besagten Kamera, die einen Bildsensor (306), einen Verstärker (310) und einen Konverter (312) aufweist,

- Berechnen einer Eingangsunsicherheit σ_i(x, y) für die Kalibrierungsdaten, welche die besagte Anzahl von Kalibrierungsbildern (801) umfassen,

- Einstellen eines Satzes von Vorkompressionsparametern g,

- Anwenden einer verlustbehafteten Vorkompression gefolgt von einer verlustfreien Kompression auf die Kalibrierungsdaten,

- Dekomprimieren der Kalibrierungsdaten,

- Berechnung einer Ausgangsunsicherheit σ_o(x, y) für die Kalibrierungsdaten,

- Wiederholen der obigen Schritte unter Erhöhen von L und g bis L_max und g_max erreicht sind,

- Auftragen des Informationsverlusts als Funktion des Satzes von Vorkompressionsparametern g,

wobei

L(x, y) die Lichtstrahlung am Punkt (x, y) ist,

g der Satz von Vorkompressionsparametern ist,

σ_i(x, y) die Unsicherheit ist, die mit der Meßschätzung der Lichtstrahlung L(x, y) am Punkt (x, y) vor der Kompression verbunden ist,

σ_o(x, y) die Unsicherheit ist, die mit der Meßschätzung der Lichtstrahlung L(x, y) am Punkt (x, y) nach Kompression und Dekompression verbunden ist.

5. Datenkompressionsverfahren nach einem der Ansprüche 1 bis 4, dadurch gekennzeichnet, daß die verlustbehaftete Vorkompression der Bilddaten im natürlichen, nicht transformierten Raum der Bilddaten angewendet wird.

6. Datendekompressionsverfahren umfassend

- Erhalten komprimierter Bilddaten, die durch das Kompressionsverfahren nach einem der Ansprüche 1 bis 5 komprimiert wurden,

- Erhalten einer den komprimierten Bilddaten zugeordneten und durch das Kompressionsverfahren nach einem der Ansprüche 1 bis 5 gespeicherten oder übertragenen Informationsverlustgrenze,

- Dekomprimieren der komprimierten Bilddaten,

- Bestimmen einer Unsicherheit für die dekomprimierten Bilddaten gemäß der folgenden Beziehung

wobei

Δ(g, L) die während des Systemkalibrierungsschritts des Kompressionsverfahrens nach einem der Ansprüche 1 bis 5 erhaltene relative Zunahme der Unsicherheit als Funktion der auf das Pixel auftreffenden Lichtstrahlung L und des Satzes von Vorkompressionsparametern g ist,

σ_o(x, y) die Unsicherheit ist, die mit der Meßschätzung der Lichtstrahlung L(x, y) am Punkt (x, y) nach Kompression und Dekompression verbunden ist,

σ_i(x, y) die während des Systemkalibrierungsschritts des Kompressionsverfahrens nach einem der Ansprüche 1 bis 5 erhaltene Unsicherheit ist, die mit der Meßschätzung der Lichtstrahlung L(x, y) am Punkt (x, y) vor der Kompression verbunden ist,

L_o(x, y) die Schätzung der Lichtstrahlung L(x, y) ist, wie sie mittels Messung nach Kompression und Dekompression abgeleitet wird,

L_i(x, y) die während des Systemkalibrierungsschritts des Kompressionsverfahrens nach einem der Ansprüche 1 bis 5 erhaltene Schätzung der Lichtstrahlung L(x, y) ist, wie sie mittels Messung vor der Kompression abgeleitet wird,

g der während des Systemkalibrierungsschritts des Kompressionsverfahrens nach einem der Ansprüche 1 bis 5 erhaltene Satz von Vorkompressionsparametern ist.

7. Datendekompressionsverfahren nach Anspruch 6, dadurch gekennzeichnet, daß g gleich 2 und Δ(g, L_o) gleich 1,15 ist.

8. Datenverarbeitungsverfahren, umfassend das Kompressionsverfahren nach einem der Ansprüche 1 bis 5 gefolgt von dem Dekompressionsverfahren nach einem der Ansprüche 6 bis 7.

9. Computerprogramm, das auf einem Computer oder einem zweckbestimmten Hardwaregerät ausgeführt wird, beispielsweise einem FPGA, das geeignet ist, um das Datenverarbeitungsverfahren nach Anspruch 8 auszuführen.

Revendications

1. Procédé de compression de données dans lequel lesdites données comprennent du bruit et de l'information, comprenant

- une étape d'acquisition de données d'image effectuée par une caméra afin d'obtenir des données d'image à compresser,

- une étape de détermination de paramètres permettant de déterminer des paramètres de pré-compression qui sont liés à une perte d'information,

- une étape de compression, et

- une étape de stockage,
ladite étape de compression comprenant

- une pré-compression avec perte pour supprimer quelque peu de bruit des données d'image suivie de

- une compression sans perte pour compresser les données d'image restantes,
caractérisé en ce que

- ladite étape de détermination de paramètres comprend

∘ une étape de calibrage du système utilisant des données de calibrage comprenant un certain nombre d'images de calibrage chacune acquises par ou simulées pour ladite caméra à une radiation de lumière connue et délivrant la quantité d'information d'image qui est perdue par l'étape de compression en fonction des paramètres de pré-compression, ladite perte d'information étant représentée par l'augmentation relative de l'incertitude associée à l'estimation de mesure de la radiation de luminière desdites données d'image après ladite étape de compression et une étape de décompression par rapport à l'incertitude desdites données d'image avant ladite étape de compression, et

∘ une étape de sélection de paramètres de pré-compression permettant de choisir lesdits paramètres de pré-compression,

- ladite pré-compression avec perte étant effectuée en utilisant les paramètres de pré-compression sélectionnés lors de l'étape de sélection de paramètres de pré-compression.

2. Procédé de compression de données selon la revendication 1, caractérisé en ce que ladite étape de détermination de paramètres détermine lesdits paramètres de pré-compression en établissant un tracé de la perte d'information en fonction desdits paramètres de pré-compression.

3. Procédé de compression de données selon l'une des revendications 1 à 2, caractérisé en ce que l'étape de détermination de paramètres détermine une limite de perte d'information et des paramètres de pré-compression garantissant que la perte d'information est inférieure à ladite limite de perte d'information.

4. Procédé de compression de données selon l'une des revendications 1 à 3, caractérisé en ce que l'étape de calibrage du système comprend

- régler la radiation de lumière L(x, y) à mesurer, à partir de L=0 (800),

- acquérir un certain nombre d'images de calibrage (801) avec ladite caméra comprenant un capteur d'image (306), un amplificateur (310) et un convertisseur (312),

- calculer une incertitude d'entrée σ_i(x, y) sur les données de calibrage comprenant ledit nombre d'images de calibrage (801),

- assigner un ensemble des paramètres de pré-compression g,

- appliquer une pré-compression avec perte suivie d'une compression sans perte aux données de calibrage,

- décompresser les données de calibrage,

- calculer une incertitude de sortie σ_o(x, y) sur les données de calibrage,

- répéter les étapes ci-dessus en augmentant L et g jusqu'à ce que L_max et g_max soient atteints,

- reporter la perte d'information en fonction de l'ensemble des paramètres de pré-compression g,

où

L(x, y) est la radiation de lumière au point (x, y),

g est l'ensemble des paramètres de pré-compression,

σ_i(x, y) est l'incertitude associée à l'estimation de mesure de la radiation L(x, y) au point (x,y), avant compression,

σ_o(x, y) est l'incertitude associée à l'estimation de mesure de la radiation L(x, y) au point (x, y) , après compression et décompression.

5. Procédé de compression de données selon l'une des revendications 1 à 4, caractérisé en ce que la précompression avec perte des données d'image est appliquée dans l'espace naturel non transformé des données d'image.

6. Procédé de décompression de données comprenant

- obtenir des données d'image compressées qui ont été compressées par le procédé de compression selon l'une des revendications 1 à 5,

- obtenir une limite de perte d'information associée auxdites données d'image compressées et stockée ou transmise par le procédé de compression selon l'une des revendications 1 à 5,

- décompresser les données d'image compressées,

- déterminer une incertitude sur les données d'image décompressées selon la relation suivante

où

Δ(g, L) est l'augmentation relative de l'incertitude en fonction de la radiation de lumière L frappant le pixel et de l'ensemble des paramètres de pré-compression g, obtenue lors de l'étape de calibrage du système du procédé de compression selon l'une des revendications 1 à 5,

σ_o(x, y) est l'incertitude associée à l'estimation de mesure de la radiation L(x, y) au point (x, y) , après compression et décompression,

σ_i(x, y) est l'incertitude associée à l'estimation de mesure de la radiation L(x, y) au point (x, y) , avant compression, obtenue lors de l'étape de calibrage du système du procédé de compression selon l'une des revendications 1 à 5,

L_o(x, y) est l'estimation de la radiation L(x, y) dérivée de la mesure, après compression et décompression,

L_i(x, y) est l'estimation de radiation L(x, y) dérivée de la mesure, avant compression, obtenue lors de l'étape de calibrage du système du procédé de compression selon l'une des revendications 1 à 5,

g est l'ensemble des paramètres de pré-compression obtenu lors de l'étape de calibrage du système du procédé de compression selon l'une des revendications 1 à 5.

7. Procédé de décompression de données selon la revendication 6, caractérisé en ce que g est égale à 2 et Δ(g, L_o) est égale à 1,15.

8. Procédé de traitement de données comprenant le procédé de compression selon l'une des revendications 1 à 5 suivi du procédé de décompression selon l'une des revendications 6 à 7.

9. Programme informatique exécuté par un ordinateur ou un dispositif matériel dédié, par exemple un FPGA, adapté pour exécuter le procédé de traitement de données selon la revendication 8.

Drawing